Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method of reconstructing an N-channel audio signal (X) based on a single-channel downmix signal (Y), the method comprising: receiving, by a decorrelating section of a parametric reconstruction system, the single-channel downmix signal (Y); processing the single-channel downmix signal (Y) to output an (N−1)-channel decorrelated signal (Z), the processing including applying respective filters to the single-channel downmix signal (Y); receiving, by a dry upmix section of the parametric reconstruction system, the single-channel downmix signal (Y) and dry upmix parameters ({tilde over (C)}), the dry upmix parameters ({tilde over (C)}) coinciding with a first portion of a set of dry upmix coefficients (C); determining a second portion of the set of dry upmix coefficients (C) based on a predefined relation between the set of dry upmix coefficients (C); outputting, by the dry upmix section, a dry upmix signal (CY) computed by mapping the single-channel downmix signal (Y) linearly in accordance with the set of dry upmix coefficients (C); receiving, by a wet upmix section of the parametric reconstruction system, the (N−1)-channel decorrelated signal (Z) and a set of wet upmix parameters (({tilde over (P)}); deriving, from the set of wet upmix parameters (({tilde over (P)}), a set of wet upmix coefficients (P); outputting, by the wet upmix section, a wet upmix signal (PZ) computed by mapping the (N−1)-channel decorrelated signal (Z) and the set of wet upmix coefficients (P); and combining, by a combining section of the parametric reconstruction system, the dry upmix signal (CY) and the wet upmix signal (PZ) to obtain a multidimensional reconstructed signal ({circumflex over (X)}) corresponding to the N-channel audio signal (X) to be reconstructed, wherein the parametric reconstruction system includes one or more processors.
This invention relates to audio signal processing, specifically reconstructing a multi-channel audio signal from a single-channel downmix signal. The problem addressed is the efficient and high-quality reconstruction of an N-channel audio signal (X) from a single-channel downmix signal (Y) using parametric techniques. The method involves a parametric reconstruction system with three main sections: a decorrelating section, a dry upmix section, and a wet upmix section. The decorrelating section processes the single-channel downmix signal (Y) to generate an (N−1)-channel decorrelated signal (Z) by applying respective filters. The dry upmix section receives the downmix signal (Y) and a subset of dry upmix parameters, then determines the remaining dry upmix coefficients based on a predefined relation. It outputs a dry upmix signal (CY) by linearly mapping the downmix signal (Y) using the full set of dry upmix coefficients (C). The wet upmix section receives the decorrelated signal (Z) and wet upmix parameters, derives the full set of wet upmix coefficients (P), and outputs a wet upmix signal (PZ) by mapping the decorrelated signal (Z) using these coefficients. Finally, the combining section merges the dry upmix signal (CY) and the wet upmix signal (PZ) to produce the reconstructed N-channel audio signal ({circumflex over (X)}). The system is implemented using one or more processors. This approach enables efficient multi-channel audio reconstruction while maintaining perceptual quality.
2. The method of claim 1 , comprising: populating an intermediate matrix having more elements than the number of received wet upmix parameters, based on the received wet upmix parameters, the intermediate matrix belonging to a predefined matrix class.
This invention relates to audio signal processing, specifically methods for generating an intermediate matrix used in audio upmixing. The problem addressed is efficiently expanding a set of received wet upmix parameters into a larger intermediate matrix while ensuring the matrix belongs to a predefined class, which likely imposes structural or mathematical constraints to maintain audio quality during upmixing. The method involves receiving a set of wet upmix parameters, which are likely coefficients or control signals derived from an audio processing algorithm. These parameters are used to populate an intermediate matrix that has more elements than the number of received parameters. The intermediate matrix is constructed such that it belongs to a predefined matrix class, which may include properties like orthogonality, symmetry, or specific spectral characteristics to ensure stable and high-quality audio rendering. The predefined class ensures that the matrix maintains desirable properties for subsequent audio processing steps, such as spatialization or channel separation. This approach allows for flexible and efficient audio upmixing by leveraging a compact set of parameters to generate a larger, structured matrix, reducing computational overhead while preserving audio fidelity. The predefined matrix class ensures that the generated matrix adheres to constraints that prevent artifacts or degradation in the upmixed audio signal. This technique is particularly useful in real-time audio applications where computational efficiency and audio quality are critical.
3. The method of claim 2 , wherein deriving the set of wet upmix coefficients comprises multiplying the intermediate matrix by a predefined matrix, wherein the set of wet upmix coefficients corresponds to a matrix resulting from the multiplication and includes more coefficients than the number of elements in the intermediate matrix.
This invention relates to audio signal processing, specifically techniques for deriving upmix coefficients in multi-channel audio systems. The problem addressed is the efficient and accurate generation of upmix coefficients that enable the conversion of a lower-channel audio signal (e.g., stereo) into a higher-channel output (e.g., surround sound) while preserving audio quality and spatial characteristics. The method involves generating an intermediate matrix from input audio signals, which represents a preliminary set of relationships between the input channels. To derive the final set of wet upmix coefficients, this intermediate matrix is multiplied by a predefined matrix. The resulting product is a matrix of upmix coefficients that has more elements than the original intermediate matrix, allowing for enhanced channel separation and spatial rendering in the output audio. The predefined matrix is designed to optimize the upmix process, ensuring that the derived coefficients effectively distribute audio energy across the expanded channel configuration while maintaining phase coherence and minimizing artifacts. This approach improves upon traditional upmix techniques by dynamically adjusting the coefficient set based on the input signal characteristics, leading to more natural and immersive audio reproduction. The method is particularly useful in applications such as home theater systems, virtual reality audio, and broadcast audio processing.
4. The method of claim 3 , wherein the predefined matrix class is one of: lower or upper triangular matrices with predefined matrix elements being zero; symmetric matrices with predefined matrix elements being equal; and products of an orthogonal matrix and a diagonal matrix with known relations between predefined matrix elements.
This invention relates to computational methods for handling specific types of matrices in numerical computations or linear algebra applications. The problem addressed involves efficiently processing matrices with predefined structural properties, such as zero elements in specific positions, symmetric elements, or relationships between elements, to optimize computational efficiency and accuracy. The method involves selecting a predefined matrix class from a set of structured matrix types. One class includes lower or upper triangular matrices, where elements outside the triangular region are zero, reducing computational complexity by eliminating unnecessary operations on zero-valued elements. Another class consists of symmetric matrices, where elements are equal across the diagonal, allowing for simplified calculations by leveraging symmetry. A third class involves products of orthogonal and diagonal matrices, where predefined relationships between elements enable efficient decomposition or inversion without full matrix operations. The method processes these matrices by applying computational techniques tailored to their structural properties, such as exploiting zero elements to skip operations, using symmetry to reduce storage or computation, or leveraging known relationships between elements to simplify matrix operations. This approach enhances performance in applications like linear system solving, matrix inversion, or eigenvalue computations by reducing computational overhead and improving numerical stability. The invention is particularly useful in fields requiring high-performance matrix computations, such as scientific computing, machine learning, or signal processing.
5. The method of claim 3 , wherein the wet upmix parameters include N(N−1)/2 wet upmix parameters, wherein populating the intermediate matrix includes obtaining values for (N−1) 2 matrix elements based on the N(N−1)/2 wet upmix parameters and knowing that the intermediate matrix belongs to the predefined matrix class, wherein the predefined matrix includes N(N−1) elements, and wherein the set of wet upmix coefficients includes N(N−1) coefficients.
This invention relates to audio signal processing, specifically methods for generating upmix parameters in multi-channel audio systems. The problem addressed is efficiently determining wet upmix parameters for converting lower-order audio signals (e.g., stereo) into higher-order formats (e.g., surround sound) while maintaining perceptual quality and computational efficiency. The method involves creating an intermediate matrix with N(N−1) elements, where N represents the number of output channels. The intermediate matrix belongs to a predefined matrix class that enforces certain structural constraints. The wet upmix parameters consist of N(N−1)/2 coefficients, which are used to populate (N−1) of the matrix elements. The remaining elements are derived based on the known matrix class properties, ensuring the matrix remains valid for audio processing. This approach reduces the number of independent parameters needed, simplifying computation while preserving audio quality. The predefined matrix class ensures that the intermediate matrix remains invertible and suitable for real-time audio processing applications. The method is particularly useful in systems where computational resources are limited, such as consumer electronics or embedded audio devices. By leveraging the matrix class constraints, the system avoids redundant calculations, improving efficiency without sacrificing audio fidelity.
6. The method of claim 2 , wherein populating the intermediate matrix includes employing the received wet upmix parameters as elements in the intermediate matrix.
This invention relates to audio signal processing, specifically methods for generating an intermediate matrix used in audio upmixing. The problem addressed involves efficiently incorporating wet upmix parameters into an intermediate matrix to enhance audio rendering quality. Wet upmix parameters are derived from audio signals and are used to adjust the spatial characteristics of the audio output, such as directionality and reverberation. The method involves receiving wet upmix parameters, which are numerical values representing spatial audio adjustments. These parameters are then directly inserted as elements within an intermediate matrix. The intermediate matrix serves as a processing stage that transforms input audio signals into a format suitable for multi-channel output, such as surround sound or binaural audio. By embedding the wet upmix parameters into the matrix, the system can dynamically adjust the audio spatialization in real-time, improving the listener's perception of sound sources. The intermediate matrix may also include other components, such as dry upmix parameters or filter coefficients, which contribute to the overall audio processing. The wet upmix parameters specifically modify the matrix to enhance the wet or reverberant portions of the audio, ensuring a balanced and immersive listening experience. This approach optimizes computational efficiency while maintaining high-quality audio output. The method is particularly useful in applications like virtual reality, gaming, and home theater systems where precise spatial audio is critical.
7. An audio decoding system comprising a first parametric reconstruction section configured to reconstruct an N-channel audio signal (X) based on a first single-channel downmix signal (Y), the audio decoding system comprising: a first decorrelating section of a parametric reconstruction system configured to perform operations comprising: receiving the single-channel downmix signal (Y); processing the single-channel downmix signal (Y), the processing including applying respective filters to the single-channel downmix signal (Y); and output an (N−1)-channel decorrelated signal (Z); a first dry upmix section configured to perform operations comprising: receiving the single-channel downmix signal (Y) and dry upmix parameters ({tilde over (C)}), the dry upmix parameters ({tilde over (C)}) coinciding with a first portion of a set of dry upmix coefficients (C); determining a second portion of the set of dry upmix coefficients (C) based on a predefined relation between the set of dry upmix coefficients (C); and outputting a dry upmix signal (CY) computed by mapping the single-channel downmix signal (Y) linearly in accordance with the set of dry upmix coefficients (C); a first wet upmix section of the parametric reconstruction system configured to perform operations comprising: receiving the (N-1)-channel decorrelated signal (Z) and a set of wet upmix parameters (({tilde over (P)}); deriving, from the set of wet upmix parameters ({tilde over (P)}), a set of wet upmix coefficients (P); and outputting a wet upmix signal (PZ) computed by mapping the (N−1)-channel decorrelated signal (Z) and the set of wet upmix coefficients (P); and a combining section configured to perform operations comprising: combining the dry upmix signal (CY) and the wet upmix signal (PZ) to obtain a multidimensional reconstructed signal ({circumflex over (X)}) corresponding to the N-channel audio signal to be reconstructed, wherein the first parametric reconstruction section includes one or more processors.
This invention relates to audio decoding systems designed to reconstruct multi-channel audio signals from a single-channel downmix signal. The system addresses the challenge of efficiently decoding spatial audio information from compressed or downmixed audio streams while preserving perceptual quality. The audio decoding system includes a parametric reconstruction section that processes a single-channel downmix signal to generate an N-channel audio output. A decorrelating section processes the downmix signal by applying filters to produce an (N−1)-channel decorrelated signal. A dry upmix section receives the downmix signal and dry upmix parameters, which form a subset of a full set of dry upmix coefficients. The system determines the remaining coefficients based on predefined relationships and outputs a dry upmix signal by linearly mapping the downmix signal using these coefficients. A wet upmix section processes the decorrelated signal using wet upmix parameters to derive wet upmix coefficients and outputs a wet upmix signal. The system combines the dry and wet upmix signals to reconstruct the original N-channel audio. The parametric reconstruction section operates using one or more processors, enabling efficient real-time decoding. This approach improves audio quality by leveraging both direct and decorrelated signal components while minimizing computational complexity.
8. The system of claim 7 , wherein the first parametric reconstruction section is configured to perform operations comprising: populating an intermediate matrix having more elements than the number of received wet upmix parameters, based on the received wet upmix parameters, the intermediate matrix belonging to a predefined matrix class.
This invention relates to audio signal processing, specifically systems for reconstructing audio signals from compressed or reduced parameter sets. The problem addressed is the efficient reconstruction of high-quality audio signals from a limited set of parameters, particularly in applications like audio coding, upmixing, or spatial audio rendering where computational efficiency and memory usage are critical. The system includes a parametric reconstruction section that processes received wet upmix parameters to generate an intermediate matrix. This matrix has more elements than the number of received parameters, meaning the system expands the input data into a larger representation. The intermediate matrix belongs to a predefined class, implying it follows specific structural or mathematical constraints to ensure stability, realism, or computational efficiency during reconstruction. The predefined matrix class may include properties like symmetry, sparsity, or adherence to perceptual audio models to optimize the reconstruction process. The system likely operates in a multi-stage pipeline where the intermediate matrix is further processed to produce the final audio output. The expansion of parameters into a larger matrix allows for more detailed or flexible audio reconstruction while maintaining computational efficiency by leveraging the predefined structure. This approach is useful in scenarios where low-latency or real-time processing is required, such as in streaming, virtual reality, or teleconferencing applications. The use of a predefined matrix class ensures that the reconstruction remains robust and computationally tractable.
9. The system of claim 8 , wherein deriving the set of wet upmix coefficients comprises multiplying the intermediate matrix by a predefined matrix, wherein the set of wet upmix coefficients corresponds to a matrix resulting from the multiplication and includes more coefficients than the number of elements in the intermediate matrix.
This invention relates to audio signal processing, specifically a system for deriving wet upmix coefficients in a multi-channel audio environment. The problem addressed is the efficient and accurate generation of upmix coefficients that transform a lower-dimensional audio signal into a higher-dimensional output while preserving spatial and perceptual audio quality. The system processes an intermediate matrix, which is derived from an input audio signal and represents a compact representation of the audio data. To generate the wet upmix coefficients, the intermediate matrix is multiplied by a predefined matrix. The resulting product is a set of coefficients that expands the dimensionality of the audio signal, producing more coefficients than the original elements in the intermediate matrix. This expansion allows for precise control over the spatial distribution of audio channels in the output, enhancing the listener's perception of sound directionality and immersion. The predefined matrix is designed to optimize the upmix process, ensuring that the derived coefficients maintain the integrity of the original audio while enabling seamless integration into multi-channel audio systems. This approach improves upon traditional methods by reducing computational complexity while maintaining high-quality audio reproduction. The system is particularly useful in applications such as virtual reality, surround sound systems, and immersive audio experiences where accurate spatial rendering is critical.
10. The system of claim 9 , wherein the predefined matrix class is one of: lower or upper triangular matrices with predefined matrix elements being zero; symmetric matrices with predefined matrix elements being equal; and products of an orthogonal matrix and a diagonal matrix with known relations between predefined matrix elements.
This invention relates to a computational system for processing matrices with predefined structural properties. The system is designed to efficiently handle matrices that exhibit specific patterns or symmetries, such as lower or upper triangular matrices where certain elements are zero, symmetric matrices where corresponding elements are equal, or matrices that are products of an orthogonal matrix and a diagonal matrix with known relationships between elements. The system leverages these predefined structures to optimize computational tasks, such as matrix multiplication, inversion, or decomposition, by reducing the number of required operations. This approach improves efficiency in applications like linear algebra computations, numerical simulations, and machine learning, where matrix operations are frequently performed. The system may include components for identifying the matrix class, applying structural constraints during computations, and validating results against expected properties. By exploiting known matrix structures, the system minimizes redundant calculations and enhances performance in high-dimensional data processing tasks.
11. The system of claim 9 , wherein the wet upmix parameters include N(N−1)/2 wet upmix parameters, wherein populating the intermediate matrix includes obtaining values for (N−1) 2 matrix elements based on the N(N−1)/2 wet upmix parameters and knowing that the intermediate matrix belongs to the predefined matrix class, wherein the predefined matrix includes N(N−1) elements, and wherein the set of wet upmix coefficients includes N(N−1) coefficients.
This invention relates to audio signal processing, specifically to systems for generating wet upmix parameters in multi-channel audio upmixing. The problem addressed is efficiently determining wet upmix parameters for converting a lower-channel audio signal (e.g., stereo) into a higher-channel output (e.g., 5.1 surround) while maintaining perceptual audio quality. The system processes an intermediate matrix that belongs to a predefined matrix class, which ensures the matrix meets certain mathematical constraints for stable audio processing. The matrix contains N(N−1) elements, where N is the number of output channels. The wet upmix parameters consist of N(N−1)/2 values, which are used to populate (N−1) 2 matrix elements. The remaining elements are derived from the predefined matrix class properties. This approach reduces computational complexity by leveraging matrix symmetry and constraints, ensuring efficient and accurate audio upmixing. The system dynamically adjusts these parameters to optimize the spatial distribution of audio signals across multiple channels, enhancing immersive listening experiences. The method ensures that the intermediate matrix remains within the predefined class, preventing artifacts and maintaining audio coherence. This technique is particularly useful in real-time audio processing applications where computational efficiency and high-quality output are critical.
12. The system of claim 8 , wherein populating the intermediate matrix includes employing the received wet upmix parameters as elements in the intermediate matrix.
This invention relates to audio processing systems, specifically for upmixing audio signals from a lower channel count to a higher channel count. The problem addressed is the need for efficient and accurate audio upmixing, particularly in scenarios where wet upmix parameters—parameters derived from processing the input audio—are used to enhance the quality of the upmixed output. The system includes an input interface for receiving an audio signal with a lower channel count and a processing module that generates an intermediate matrix. This intermediate matrix is used to transform the input audio into an output audio signal with a higher channel count. The key innovation involves populating the intermediate matrix using wet upmix parameters, which are dynamically adjusted based on the input audio characteristics. These parameters are applied as elements within the intermediate matrix to ensure that the upmixing process preserves spatial and spectral details of the original audio. The system may also include a parameter generation module that derives the wet upmix parameters from the input audio, ensuring that the upmixing process is adaptive and context-aware. The intermediate matrix, populated with these parameters, is then applied to the input audio to produce the final upmixed output. This approach improves the naturalness and clarity of the upmixed audio compared to static or less adaptive methods. The system is particularly useful in applications such as surround sound processing, virtual reality audio, and multi-channel audio reproduction.
13. A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations of reconstructing an N-channel audio signal (X) based on a single-channel downmix signal (Y), the operations comprising: receiving, by a decorrelating section of a parametric reconstruction system, the single-channel downmix signal (Y); processing the single-channel downmix signal (Y) to output an (N−1)-channel decorrelated signal (Z), the processing including applying respective filters to the single-channel downmix signal (Y); receiving, by a dry upmix section of the parametric reconstruction system, the single-channel downmix signal (Y) and dry upmix parameters ({tilde over (C)}), the dry upmix parameters ({tilde over (C)}) coinciding with a first portion of a set of dry upmix coefficients (C); determining a second portion of the set of dry upmix coefficients (C) based on a predefined relation between the set of dry upmix coefficients (C); outputting, by the dry upmix section, a dry upmix signal (CY) computed by mapping the single-channel downmix signal (Y) linearly in accordance with the set of dry upmix coefficients (C); receiving, by a wet upmix section of the parametric reconstruction system, the (N−1)-channel decorrelated signal (Z) and a set of wet upmix parameters (({tilde over (P)}); deriving, from the set of wet upmix parameters (({tilde over (P)}), a set of wet upmix coefficients (P); outputting, by the wet upmix section, a wet upmix signal (PZ) computed by mapping the (N−1)-channel decorrelated signal (Z) and the set of wet upmix coefficients (P); and combining, by a combining section of the parametric reconstruction system, the dry upmix signal (CY) and the wet upmix signal (PZ) to obtain a multidimensional reconstructed signal ({circumflex over (X)}) corresponding to the N-channel audio signal (X) to be reconstructed.
The invention relates to audio signal processing, specifically reconstructing an N-channel audio signal from a single-channel downmix signal. The problem addressed is efficiently generating a multi-channel audio output from a compressed or reduced-channel input while preserving spatial audio quality. The system processes the downmix signal through three main sections: a decorrelating section, a dry upmix section, and a wet upmix section. The decorrelating section applies filters to the downmix signal to produce an (N−1)-channel decorrelated signal. The dry upmix section uses a subset of dry upmix parameters to compute a full set of dry upmix coefficients, then linearly maps the downmix signal to generate a dry upmix signal. The wet upmix section derives wet upmix coefficients from wet upmix parameters and maps the decorrelated signal to produce a wet upmix signal. The system combines these signals to reconstruct the original N-channel audio. The method optimizes computational efficiency by deriving partial coefficients and leveraging predefined relations between parameters, ensuring high-quality spatial audio reconstruction from a single-channel input.
Unknown
April 7, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.